Measuring health behaviours with electronic healthcare data: the case of medication adherence

Abstract

Background: Electronic healthcare data (EHD) are a rich and accessible source of information on routine clinical care and can provide valuable insights into patient and clinician behaviours. Yet, although EHD are increasingly used in medical research, they are less considered in health psychology due to lower awareness of their potential to measure behaviour, and lack of accessible software and algorithms to facilitate such analyses. Methods: We developed AdhereR, a package for the statistical environment R, to facilitate computing EHD-based estimates of medication adherence. Several functions for calculating medication persistence (treatment episodes) and implementation (Continuous Medication Availability; CMA) and visualising medication histories were implemented based on a review of current adherence guidelines, definitions and operationalisations. We illustrate the use of AdhereR with an example dataset of 2-year records of 100 patients and describe the various analysis choices possible and how they can be adapted to different health conditions and types of medications. Results: Nine CMA variations can be computed for single intervals, multiple treatment episodes, and sliding observation windows. Different parameter choices result in different estimates (e.g. median 1-year implementation range 56%-83% in one simulation comparing single-interval CMAs). Separating persistence from implementation resulted in higher CMA estimates (e.g. median increased from 57% to 85% when comparing single with per-episode CMA). Discussion: AdhereR facilitates transparent and replicable calculations of EHD-based medication adherence. Analysis choices impact calculations and need careful consideration. As recruiting and maintaining participation in research becomes increasingly difficult, health psychologists need to consider exploiting the potential of routinely-collected healthcare data.